Let's build an AI agent
summary
Live-build of an agent that adds Gemini via Genai SDK chat memory, function-calling tools, and a tool loop that re-invokes the model until goals are met; uses AstroDB RAG for document search, a date tool, a simple add function, and MCP servers like Tavly and Runno to fetch web data and run code in WebAssembly. The core loop is ~20 lines that detects function_call, runs declared functions, returns results, and repeats; multi-agent chaining works (research agent calls fetch agent) and the system correctly computes primes by running Python in Runno (15th=47, 25th=97).
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:382025 and AI agents
- 3:44What is an AI agent
- 6:13AI capabilities evolution
- 10:02Text generation with Langflow
- 13:31Adding agent memory
- 18:04Defining agent tools
- 24:01Agent tool loop
- 29:23Real tools: date and RAG
- 35:08Multi-agent systems
- 39:02Model Context Protocol
- 41:08MCP server integration
- 47:04Super agents with tools
- 48:08Agent design considerations
related talks
50mLet's build an AI agent
39mBuild the new generation of AI Agents
01h 48mBuilding Conversational Agents
05h 56mFrom the main stage to the terminal
22mClient side Web AI Agents: Building smarter user experiences for future agentic ...
30mLive Demo – Develop and Deploy a Trustworthy Multi-Agent System
10mBuilding intelligent agents with Gemma 3
31mHow Web Apps Are Evolving into AI Agents